# Too high load and QPS levels in QAN MySQL

**URL:** <https://forums.percona.com/t/too-high-load-and-qps-levels-in-qan-mysql/39417>\
**Category:** PMM 3.x\
**Tags:** pmm, mysql\
**Created:** [September 19, 2025, 7:59am UTC](https://forums.percona.com/t/too-high-load-and-qps-levels-in-qan-mysql/39417 "2025-09-19T07:59:43Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![svist](https://avatars.discourse-cdn.com/v4/letter/s/8797f3/32.png) [@svist](https://forums.percona.com/u/svist)\
**Post date:** [September 19, 2025, 7:59am UTC](https://forums.percona.com/t/too-high-load-and-qps-levels-in-qan-mysql/39417/1 "2025-09-19T07:59:43Z")

</div>

MySQL: Server version: 5.7.35-38-log Percona Server (GPL), Release ‘38’, Revision ‘3692a61’

mysql\> show variables like ‘%slow%’;  
`±----------------------------------±-----------------------------------+`  
`| Variable_name | Value |`  
`±----------------------------------±-----------------------------------+`  
`| log_slow_admin_statements | OFF |`  
`| log_slow_filter | |`  
`| log_slow_rate_limit | 1 |`  
`| log_slow_rate_type | session |`  
`| log_slow_slave_statements | OFF |`  
`| log_slow_sp_statements | ON |`  
`| log_slow_verbosity | |`  
`| max_slowlog_files | 0 |`  
`| max_slowlog_size | 0 |`  
`| slow_launch_time | 2 |`  
`| slow_query_log | OFF |`  
`| slow_query_log_always_write_time | 10.000000 |`  
`| slow_query_log_file | /var/lib/mysql/db-mysql-1-slow.log |`  
`| slow_query_log_use_global_control | |`  
`±----------------------------------±-----------------------------------+`

PMM: 3.3.1

QAN:  
Query Count: 45.37m QPS  
Rows Sent: 45.31m (per sec)  
This is nonsense.

We have absurdly high query counts and load level in QAN.  
Real load: 10k-30k.  
Previous issue: performance\_schema.events\_statements\_summary\_by\_digest table overflowed, stuck at 10k records for six months (performance\_schema\_digests\_size = 10k).  
Fixed by increasing performance\_schema\_digests\_size and restarting the server (table cleared).  
Since then, QAN load gradually rises to unrealistic levels.

---

<div class="post-metadata">

**Author:** ![nurlan](https://sea1.discourse-cdn.com/flex019/user_avatar/forums.percona.com/nurlan/32/1523_2.png) [@nurlan](https://forums.percona.com/u/nurlan)\
**Post date:** [September 20, 2025, 8:52am UTC](https://forums.percona.com/t/too-high-load-and-qps-levels-in-qan-mysql/39417/2 "2025-09-20T08:52:02Z")

</div>

Hi, have you restart pmm-agent after increasing performance\_schema\_digests\_size? It might be caused by pmm-agents internal cache size being lower then MySQL. pmm-agent reads performance\_schema\_digests\_size only once at the beginning. Restart of pmm-agent should help.

---

<div class="post-metadata">

**Author:** ![svist](https://avatars.discourse-cdn.com/v4/letter/s/8797f3/32.png) [@svist](https://forums.percona.com/u/svist)\
**Post date:** [September 22, 2025, 7:01am UTC](https://forums.percona.com/t/too-high-load-and-qps-levels-in-qan-mysql/39417/3 "2025-09-22T07:01:28Z")

</div>

Yes, the agent was restarted. The issue isn’t that we’re missing queries in PMM like before - that was fixed. The problem now is we’re seeing wild metrics on query count and frequency.

 ![pmm](https://us1.discourse-cdn.com/flex019/uploads/percona1/original/3X/7/b/7bfe3f5f16f6214e5bfe8e7824682970df9843db.png)

and real status (from zabbix)

 ![image](https://us1.discourse-cdn.com/flex019/uploads/percona1/original/3X/c/2/c22d9b1f3325cf83073cd4c636de31dcf82faa73.png)